add another test for corrmm

theano/tensor/nnet/tests/test_corr.py

@@ -427,6 +427,35 @@ class TestGroupCorr2d(Grouped_conv_noOptim):
     conv2d_gradi = staticmethod(corr.CorrMM_gradInputs)
     flip_filter = True
 
+    def test_graph(self):
+        # define common values  first
+        groups = 3
+        bottom = np.random.rand(3, 6, 5, 5)
+        kern = np.random.rand(9, 2, 3, 3)
+        bottom_sym = T.tensor4('bottom')
+        kern_sym = T.tensor4('kern')
+
+        # grouped convolution graph
+        conv_group = self.conv2d(num_groups=groups)(bottom_sym, kern_sym)
+        gconv_func = theano.function([bottom_sym, kern_sym], conv_group)
+
+        # Graph for the normal hard way
+        kern_offset = kern_sym.shape[0] // groups
+        bottom_offset = bottom_sym.shape[1] // groups
+        split_conv_output = [self.conv2d()(bottom_sym[:, i * bottom_offset:(i + 1) * bottom_offset, :, :],
+                             kern_sym[i * kern_offset:(i + 1) * kern_offset, :, :, :])
+                             for i in range(groups)]
+        concatenated_output = T.concatenate(split_conv_output, axis=1)
+        conv_func = theano.function([bottom_sym, kern_sym], concatenated_output)
+
+        # calculate outputs for each graph
+        gconv_output = gconv_func(bottom, kern)
+        conv_output = conv_func(bottom, kern)
+
+        # compare values
+        utt.assert_allclose(gconv_output, conv_output)
+
+
 if __name__ == '__main__':
 
     t = TestCorr2D('setUp')